U.S. export controls on Nvidia’s advanced AI chips are reaching beyond direct sales to China. The Biden administration has introduced new approval requirements for exports to certain Middle Eastern buyers, citing concerns that the chips could ultimately reach China.
The move highlights how the supply of computing power has become a focus of the technology competition between the U.S. and China. It also puts scrutiny on the path chips take after they are sold.
Why Middle East exports are under review
Nvidia must now obtain U.S. government approval before exporting its high-performance H100 and A100 chips to certain buyers in the Middle East. The countries covered by the restrictions have not been named.
According to a report in The Guardian, U.S. officials are concerned that countries such as Saudi Arabia and the United Arab Emirates could resell the chips to China. The concern is not limited to a direct shipment: a sale to a buyer in one country could still provide China with access if the hardware is transferred onward.
The source says the two Gulf states have recently expanded their AI ties with China and ordered thousands of Nvidia H100 chips. It also reports that Chinese researchers have a strong presence at King Abdullah University of Science and Technology (Kaust) in Saudi Arabia.
The United Arab Emirates has also been identified by the U.S. as a potential “transshipment point” used by Russia to evade sanctions, according to the report. That history adds to U.S. scrutiny of whether controlled technology could pass through the region to another destination.
Concerns extend to AI training
U.S. officials are also concerned about how computing resources might be used after they leave the country. A senior U.S. trade lawyer told The Guardian that, amid a shortage in China, state-affiliated Chinese companies could train AI systems abroad.
This possibility broadens the policy question. Export controls can focus on where a chip is shipped, but officials are also weighing whether access to computing power outside China could still help Chinese organizations develop AI systems.
Advanced chips such as Nvidia’s are used to train sophisticated models, including ChatGPT. As AI grows more important to the economy and national security, control over the hardware needed for training has become part of the broader U.S.-China competition.
Controls on Nvidia chips may widen
The new Middle East approval requirement comes after the Biden administration banned sales of Nvidia’s top-of-the-line chips to China and Russia over security concerns. Nvidia responded by developing less powerful versions that could be sold in China.
The U.S. government is reportedly planning to ban those versions as well. If that happens, the pressure on chip makers and buyers would extend beyond the most powerful products to include hardware designed for markets where the top models are restricted.
Together, the measures show why export policy is increasingly concerned with the full route and use of AI chips. U.S. officials are trying to limit Beijing’s access to critical technology, while also assessing whether buyers abroad could offer another channel to computing capacity.
What the restrictions mean
For Nvidia, the rules add a government review step for some Middle Eastern exports of its H100 and A100 chips. For prospective buyers, the restrictions mean that access to advanced AI hardware may depend not only on the destination, but also on concerns about resale and the organizations that could benefit from the equipment.
The article does not identify the countries or buyers covered by the new requirements. It does, however, describe a policy direction: as the race for AI computing power intensifies, U.S. controls are being considered across both direct exports and possible routes through third countries.